DataCenterGame/app/etl/dw/dw_image_cost_day.py

from logging import handlers
from model.DataBaseUtils import MysqlUtils, CkUtils
from model.DateUtils import DateUtils
from model.DingTalkUtils import DingTalkUtils
from datetime import datetime
import logging

# logging.getLogger().setLevel(logging.WARNING)


db = MysqlUtils()
ck = CkUtils()
du = DateUtils()


def run(dt):
    sql = f"""
            SELECT a.dt,b.type,count(*) as ct,sum(a.cost),sum(view_count),sum(click_count),sum(follow_count),sum(order_count),sum(order_amount),
            title,description,book,platform,stage,e.channel,pitcher,ifnull(image_id,''),
            g.last_modified_time,g.campaign_id 
            from 
            ad_cost_day a 
            left join ad_info b on a.ad_id=b.ad_id
            left join adcreative_info c on b.adcreative_id=c.adcreative_id
            left join channel_by_account_daily e on b.account_id=e.account_id and a.dt=e.dt
            left join channel_info_daily f on e.channel=f.channel and e.dt=f.dt
            left join campaign_info g on b.campaign_id = g.campaign_id 
            where a.dt='{dt}'  and c.is_video=0 and g.campaign_id is not null
            group by a.dt,b.type,title,description,book,platform,stage,image_id,e.channel,pitcher
            
            """

    # print(sql)
    data = db.quchen_text.get_data_list(sql)

    # print(data)

    # 图片链接拼接

    li = []
    for i in data:
        # print(i)
        li.extend(i[16].split(','))
    # TODO:之后如果一天产生的图片过多,可能超过sql的字符限制
    # 之后数据使用hive,来进行数据存储

    sql3 = f"select image_id,preview_url,signature,width,height,size,`type` from image_info where  image_id in ({str(set(li))[1:-1]})"

    image_di = {}
    image_data = db.quchen_text.getData(sql3)
    signature_dict = {}  # key signature_id  v:(pitcher,last_modified_time)
    for x in image_data:
        image_di[x[0]] = (x[1], x[2], x[3], x[4], x[5], x[6])

    for i in data:
        signature_tmp = ''
        for j in i[16].split(','):
            signature_tmp = signature_tmp + ',' + (image_di.get(j)[1] if image_di.get(j) else ' ')
        signature_tmp = signature_tmp[1:]
        if signature_tmp not in signature_dict.keys():
            signature_dict[signature_tmp] = (i[15], i[17], i[18])
        else:
            sig_last_modified_time = signature_dict[signature_tmp][1]
            if sig_last_modified_time is None:
                signature_dict[signature_tmp] = (i[15], i[17], i[18])
            elif i[17] is not None and i[17] < sig_last_modified_time:
                signature_dict[signature_tmp] = (i[15], i[17], i[18])
    # print(image_di)

    for i in data:
        preview_url = ''
        signature = ''
        width = '0'
        height = '0'
        image_id = ''
        size = ''
        type = ''
        video_length = None
        video_bit_rate = None
        video_meta_data = None
        download_path = None
        for j in i[16].split(','):
            if image_di.get(j):
                image_id = image_id + ',' + j
                preview_url = preview_url + ',' + image_di.get(j)[0]
                signature = signature + ',' + image_di.get(j)[1]
                width = str(image_di.get(j)[2])
                height = str(image_di.get(j)[3])
                size = size + ',' + str(image_di.get(j)[4])
                type = type + ',' + str(image_di.get(j)[5])
            else:
                image_id = image_id + ',' + j
                preview_url = preview_url + ',' + ' '
                signature = signature + ',' + ' '
                width = '0'
                height = '0'
                size = size + ',' + '0'
                type = type + ',' + ' '
        signature = signature[1:]
        owner = signature_dict[signature][0]
        campaign_id = signature_dict[signature][2]
        i[16] = image_id[1:]
        i.append(preview_url[1:])
        i.append(signature)
        i.append(0)
        i.append(width)
        i.append(height)
        i.append(size[1:])
        i.append(type[1:])
        i.append(video_length)
        i.append(video_bit_rate)
        i.append(video_meta_data)
        i.append(download_path)
        i.append(campaign_id)
        i.append(owner)
    data_new = []
    for i in data:
        i = i[:17] + i[19:]
        data_new.append(i)
    data = data_new


    # exit(0)
    sql_video = f""" 
            select foo.*,foo2.pitcher as owner from
            (SELECT a.dt,b.type,count(*),sum(a.cost),sum(view_count),sum(click_count),sum(follow_count),sum(order_count),sum(order_amount),
            title,description,book,platform,stage,e.channel,pitcher,ifnull(image_id,''),g.preview_url,g.signature,1,
            g.width,g.height,g.`size` ,g.`type` as video_type ,g.video_length ,g.byte_rate ,g.video_meta_data,g.download_path
            ,min(h.last_modified_time) as last_modified_time , h.campaign_id 
            from 
            ad_cost_day a 
            left join ad_info b on a.ad_id=b.ad_id
            left join adcreative_info c on b.adcreative_id=c.adcreative_id
            left join channel_by_account_daily e on b.account_id=e.account_id and a.dt=e.dt
            left join channel_info_daily f on e.channel=f.channel and e.dt=f.dt
            left join video_info g on c.image_id=g.video_id
            left join campaign_info h on b.campaign_id = h.campaign_id 
            where a.dt='{dt}'  and c.is_video=1 and h.campaign_id  is not null
            group by a.dt,b.type,title,description,
            book,platform,stage,image_id,e.channel,pitcher) as foo
            inner join             
            (select pitcher,min(h.last_modified_time) as last_modified_time  
            from 
            ad_cost_day a 
            left join ad_info b on a.ad_id=b.ad_id
            left join adcreative_info c on b.adcreative_id=c.adcreative_id
            left join channel_by_account_daily e on b.account_id=e.account_id and a.dt=e.dt
            left join channel_info_daily f on e.channel=f.channel and e.dt=f.dt
            left join video_info g on c.image_id=g.video_id
            left join campaign_info h on b.campaign_id = h.campaign_id 
            where a.dt='{dt}'  and c.is_video=1 and h.campaign_id  is not null
            group by pitcher,h.last_modified_time ) as foo2 
            on foo.pitcher=foo2.pitcher and foo.last_modified_time=foo2.last_modified_time
            """

    data_video = db.quchen_text.get_data_list(sql_video)
    data_new = []
    for i in data_video:
        i = i[:-3] + i[-2:]
        data_new.append(i)
    data.extend(data_new)


    # 进行数据存储

    db.dm.execute(f'delete from dw_image_cost_day where dt="{dt}"')
    db.dm.executeMany(
        '''replace into dw_image_cost_day 
        (dt,type,use_times,cost,view_count,click_count,follow_count,order_count,
        order_amount,title,description,book,platform,stage,channel,pitcher,image_id,
        preview_url,signature,is_video,width,height,size,format,video_length,
        video_bit_rate,video_meta_data,download_path,campaign_id,owner)
        values (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)''',
        data)

    # # 图片数据归属权改变
    # # 大于五天的数据进行，消耗超过5000的owner置位null
    # owner_sql = '''
    #     UPDATE dw_image_cost_day
    #     set owner = null
    #     WHERE
    #     dt<date_add(now(),interval -5 day) or cost>5000
    # '''
    # db.dm.execute(owner_sql)


    # ck对应数据也保存一份
    # 1.进行当天相关的分区删除
    ck.client.execute(f''' alter table dw_image_cost_day drop partition '{dt}' ''')
    col = ['dt', 'type', 'use_times', 'cost', 'view_count', 'click_count',
           'follow_count', 'order_count', 'order_amount', 'title', 'description',
           'book', 'platform', 'stage', 'channel', 'pitcher', 'image_id', 'preview_url',
           'signature', 'is_video', 'width', 'height', 'size', 'format', 'video_length',
           'video_bit_rate', 'video_meta_data', 'download_path','campaign_id', 'owner']
    # ck存入前进行数据格式化
    for _ in data:
        # data= [col if col else 'null' for col in data ]
        _[0] = datetime.strptime(_[0], '%Y-%m-%d')
        _[1] = str(_[1]) if _[1] is not None else None
        _[2] = int(_[2]) if _[2] is not None else None
        _[3] = float(_[3]) if _[3] is not None else None
        _[4] = int(_[4]) if _[4] is not None else None
        _[5] = int(_[5]) if _[5] is not None else None
        _[6] = int(_[6]) if _[6] is not None else None
        _[7] = int(_[7]) if _[7] is not None else None
        _[8] = float(_[8]) if _[8] is not None else None
        _[19] = str(_[19])
        _[20] = str(_[20]) if _[20] is not None else None
        _[21] = str(_[21]) if _[21] is not None else None
        _[22] = str(_[22]) if _[22] is not None else None
        _[23] = str(_[23]) if _[23] is not None else None

    col_str = ','.join(col)
    logging.info('ck填充数据进入')
    ck.client.execute('SET max_partitions_per_insert_block=1000;')
    ck.client.execute('insert into dw_image_cost_day ({}) values'.format(col_str), data)
    logging.info('ck填充数据,结束')


def hourly():
    try:
        logging.info('广告数据清洗,开始')
        run(du.getNow())
        logging.info('广告数据清洗,结束')
    except Exception as e:
        logging.error(str(e))
        DingTalkUtils().send("广告数据清洗失败\n" + str(e))


def day():
    logging.info('广告数据清洗,开始')
    for i in du.getDateLists(du.get_n_days(-10), du.get_n_days(-1)):
        # print(i)
        run(i)
    logging.info('广告数据清洗,结束')


if __name__ == '__main__':
    # run('2021-05-18')

    logging.basicConfig(
        handlers=[
            logging.handlers.RotatingFileHandler('.ad_hourly.log',
                                                 maxBytes=10 * 1024 * 1024,
                                                 backupCount=5,
                                                 encoding='utf-8')
            , logging.StreamHandler()  # 供输出使用
        ],
        level=logging.INFO,
        format="%(asctime)s - %(levelname)s %(filename)s %(funcName)s %(lineno)s - %(message)s"
    )
    # -495
    #
    # for i in du.getDateLists(du.get_n_days(-495), du.get_n_days(0)):
    #     print(i)
    #     run(i)

    # print(du.get_n_days(-20))
    run(du.get_n_days(0))